Router Resource Estimator

Estimate the system resources to allocate for the Apollo Router

The router resource estimator estimates the system resources you need to allocate for the Apollo Router to run production traffic. The resource estimator can help you get a starting baseline that you can then iterate on.

Usage

To use the estimator:

Enter your estimated request rate per second and peak request rate per second.
Adjust the other parameters to match your estimated production load.
- When selecting the number of instances, select fewer for efficiency, more for safety.
Be sure to test and iterate.

Router resource estimator

Parameters

Request Rate per Second

RPS

Average number of requests per second (RPS) you expect the router to receive

Peak Request Rate per Second

RPS

The highest estimated number of requests per second (RPS) you expect the router to receive

Baseline Subgraph Latency

The baseline response time of a typical end-to-end response from your graph in milliseconds (ms)

Client Request Size

The size of a typical client request in Megabytes (MB)

Client Response Size

The size of a typical client response in Megabytes (MB)

Number of Instances

Number of router instances running in parallel

Assumptions

Base Memory

The amount of memory a router is configured with in Megabytes (MB)

Query Planner Memory

The amount of memory allocated to the query planner in Megabytes (MB)

Results

💡 TIP

Enter parameters to see results.

Based on the provided parameters and assumptions, each router instance likely requires a minimum of __ vCPUs for average traffic and a minimum of __ vCPUs for peak traffic.

We also recommend __MB of memory per router instance with a limit of __MB of memory.

Kubernetes configuration for a single instance:

1
resources:
2
  requests:
3
    memory: '__mb'
4
    cpu: '__m'
5
  limits:
6
    memory: '__mb'
7
    # no CPU limit to avoid throttling

Home